Crash on warn?

[dnil]

Apparently the automation sends log output to the equivalent of /dev/null. This is not good procedure. If this is to continue, we should proabably browse through and change a lot of dire warnings to crashes. If no-one reads them, things can go south rather quickly. Crashing early would increase pipeline compatibility and bug finding, no doubt. But be a pain for a while.

[dnil]

A strict mode? Postpone until separating out a more slowly moving parser part?

[dnil]

Let's follow up on https://github.com/Clinical-Genomics/Deviations/issues/715 and changes to the procedure. If the log is read at least once in a while on prod, and on verification testing, I think we should be fine. It does simplify day to day life if we don't crash on everything. Compare #5048 which would have been a crash as well. 🤷